Skip to content
This repository was archived by the owner on Jan 15, 2025. It is now read-only.

Comments

cross train and lu/qna build optimization for composer#889

Merged
feich-ms merged 26 commits intomasterfrom
feich/crossTrainOptimization
Jul 23, 2020
Merged

cross train and lu/qna build optimization for composer#889
feich-ms merged 26 commits intomasterfrom
feich/crossTrainOptimization

Conversation

@feich-ms
Copy link
Contributor

@feich-ms feich-ms commented Jul 13, 2020

fix #883, #880, #879, #878 and microsoft/botbuilder-dotnet#4144. All these issues are related with composer crosstrain and lu/qna:build scenarios. This PR is to resolve these issues including:

  1. add filter meta data for all qna pairs
  2. adjust the recognizer generation logic for qna to generate recognizer dialog files for each dialog instead of sharing single one of botName, but in settings file, the kb ids of qna recognizers will be same for same locale.
  3. Optimize lu/qna builder interface and flow to write out the empty crosstrained recognizer for both lu and qna files are empty scenario
  4. Do not add any crosstrained contents to lu or qna files if all lu files or all qna files are empty in crosstrained flow
  5. adjust and add Unit tests

Now composer only need to call below flow directly and DO NOT need any other tricy calling logic:
call crosstrain() -> lubuild.loadContents() ->lubuild.build() -> lubuild.writeDialogAssets() -> qnabuild.loadContents() -> qnabuild.build() -> qnabuild.writeDialogAssets()

@feich-ms feich-ms marked this pull request as ready for review July 15, 2020 06:19
@feich-ms feich-ms requested a review from munozemilio as a code owner July 15, 2020 06:19
@codecov-commenter
Copy link

codecov-commenter commented Jul 15, 2020

Codecov Report

Merging #889 into master will decrease coverage by 2.92%.
The diff coverage is 20.73%.

Impacted file tree graph

@@            Coverage Diff             @@
##           master     #889      +/-   ##
==========================================
- Coverage   60.45%   57.53%   -2.93%     
==========================================
  Files         220      219       -1     
  Lines       17887    16503    -1384     
  Branches     2316     2265      -51     
==========================================
- Hits        10814     9495    -1319     
+ Misses       6494     6428      -66     
- Partials      579      580       +1     
Impacted Files Coverage Δ
packages/lu/src/parser/lu/luMerger.js 76.34% <0.00%> (-3.77%) ⬇️
packages/lu/src/parser/lubuild/builder.ts 11.83% <0.00%> (-0.98%) ⬇️
packages/lu/src/parser/lubuild/core.ts 15.96% <ø> (-0.34%) ⬇️
packages/lu/src/parser/utils/enums/CLI-errors.js 100.00% <ø> (ø)
packages/lu/src/parser/qnabuild/builder.ts 19.86% <3.73%> (-2.87%) ⬇️
packages/lu/src/parser/qnabuild/core.ts 23.52% <50.00%> (+1.20%) ⬆️
packages/lu/src/utils/filehelper.ts 40.88% <50.00%> (-9.35%) ⬇️
packages/qnamaker/src/commands/qnamaker/build.ts 77.00% <83.33%> (+0.15%) ⬆️
packages/lu/src/parser/cross-train/crossTrainer.js 92.43% <97.61%> (-2.29%) ⬇️
packages/luis/src/commands/luis/build.ts 79.56% <100.00%> (-0.44%) ⬇️
... and 104 more

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 88a2a42...e32c259. Read the comment docs.

@boydc2014
Copy link
Contributor

@alanlong9278 @luhan2017 can you help verify this feature is as what we expected at composer side

@feich-ms feich-ms merged commit 4c45f4a into master Jul 23, 2020
@feich-ms feich-ms deleted the feich/crossTrainOptimization branch July 23, 2020 07:38
Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

crosstrain should add filter for qna only scenarios

4 participants